Cut input latency without buying anything

This is almost always the single biggest win, and it is the one most people skip because it lives in the television or monitor menu rather than in the game. Consumer displays apply picture processing by default: motion smoothing, noise reduction, dynamic contrast. Each stage buffers frames, and buffering frames is latency.

Find the picture mode menu on the display itself and select Game, or turn off every processing option individually if there is no Game preset. On a mid-range television this commonly takes the display from around 90 milliseconds of processing delay to under 20. There is no software setting on your computer or console that can undo it, which is why it goes first.

Expect: 20 to 100 ms on a television, 0 to 10 ms on a monitor that was already in a gaming preset.

When a game renders faster than the display can show frames, finished frames sit in a queue waiting their turn. Every frame in that queue is a delay between your input and what you see. Traditional vertical sync makes this worse, because it lets the queue grow.

The reliable fix is to cap the frame rate slightly below your refresh rate, typically three frames below, and leave variable refresh rate on if the display supports it. On a 120 Hz screen, cap at 117. This keeps the queue empty without reintroducing tearing. Where a game offers a low latency or anti-lag option in its graphics menu, turn it on as well.

Expect: 10 to 40 ms, and a noticeably steadier feel even when the average frame rate is unchanged.

Latency and frame rate are linked: each frame takes one frame-time to produce, so a game running at 60 frames per second cannot respond faster than roughly 16 ms per stage of its pipeline. Going from 60 to 120 halves that.

The settings that cost the most frame rate and return the least visual quality are usually shadow resolution, volumetric or global illumination quality, and screen-space reflections. Drop those three one notch before touching texture quality, which is almost free if you have the video memory for it.

Expect: 8 to 16 ms going from 60 to 120 frames per second.

Some effects are applied after the frame is otherwise finished, and a few of them need information from the previous frame to work. Motion blur and certain temporal anti-aliasing implementations fall into this group, as do most forms of frame generation.

Frame generation deserves a specific warning: it increases the number of frames you see while slightly increasing the delay before your input appears, because it has to hold a real frame back in order to interpolate. If a game feels sluggish despite a high frame counter, this is usually why.

Expect: 4 to 20 ms, depending entirely on whether frame generation was on.

Recording software, chat overlays, performance monitors and store overlays all hook into the rendering path. Individually each costs little; three at once on a modest machine is measurable, and they are also the most common cause of erratic frame times, which feel worse than a consistently slower average.

Turn them all off, confirm the improvement, then add back only the ones you actually use. Running the game in exclusive fullscreen rather than borderless windowed helps on some systems and makes no difference on others, so test it rather than assuming.

Expect: 2 to 10 ms, plus steadier frame pacing.

This is the step everyone does first and it belongs near the bottom. A modern wireless controller or mouse on a healthy connection adds a handful of milliseconds. Plugging it in with a cable removes most of that, and raising the polling rate from 125 Hz to 1000 Hz removes a few more.

It is worth doing, and it is worth doing after the display and the frame queue, because those two are where the tens of milliseconds live. A wired mouse on a television in cinema mode is still a slow chain.

Expect: 1 to 8 ms.

You do not need specialist equipment. Set a phone to record slow motion at 240 frames per second, point it at both your hand and the screen, and press a button that causes an immediate, obvious visual change. Count the recorded frames between your finger moving and the screen changing, then multiply by 1000 and divide by 240 to get milliseconds.

At 240 frames per second each frame is about 4 ms, so this method resolves to roughly that. Do it five times and take the median, because your reaction and the capture are not synchronised. Measure once before you start and once at the end, and you will know whether any of this worked instead of guessing.

Expect: an answer, which is the point of the exercise.